Social Media Investigations using Shared Photos
Speaker: Prof Danilo Montesi
Danilo Montesi is full professor of database and information systems at the Department of Computer Science and Engineering of the University of Bologna since 2005. Before joining the University of Bologna he had permanent positions with: University of Camerino, University of Milan and University of East Anglia - UK. He had visited under different grants: Department of Computer Systems and Telematics, University of Trondheim - Norway, Imperial College of London - UK, Department of Computing, Purdue University - USA, Rutherford Appleton Laboratory - UK, British Telecom Research Labs - UK and University of Lisboa - Portugal. His principal interests are in the area of database and information systems. He was vice head of the department from 2006 until 2009 and was teaching director of undergraduate and postgraduate computer science programs from 2009 until 2012. He is a faculty member of Bologna Business School since 2001.
In recent years, the spread of smartphones has attributed to changes in the user behaviour with respect to multimedia content sharing on online social networks (SNs). One noticeable behaviour is taking pictures using smartphone cameras and sharing them with friends through online social platforms. On the downside, this has contributed to the growth of the cyber crime through SNs. In this paper, we present a method to extract the characteristic fingerprint of the source camera from images being posted on SNs. We use this technique for two investigation activities i) smartphone verification: correctly verifying if a given picture has been taken by a given smartphone and ii) profile linking: matching user profiles belonging to different SNs. The method is robust enough to verify the smartphones in spite of the fact that the images get downgraded during the uploading/downloading process. Also, it is capable enough to compare different images belonging to different SNs without using the original images. We evaluate our process on real dataset using three different social networks and five different smart- phones. The results, show smartphone verification and profile linking can provide 96.48% and 99.49% respectively, on an average of the three social networks, which shows the effectiveness of our approach.